Database Reference
In-Depth Information
We've also added read access for All Authenticated Users to all of these
datasets. Normally, this entry would not be present in any of the datasets.
However, we opened up access to grant our readers the ability to look at
the data. (It wouldn't be much use for teaching people about BigQuery
otherwise.)
Tables
There are a number of tables used in our application, but we'll discuss only
one type of table here: the
logs.device_*
tables. These tables contain
the daily sensor data streams. We use the
*
wildcard on the table names
when describing them because the new tables are created every day, and the
table names include the date. For example, the February 24, 2014 debug
log table would be
logs.device_20140224
. The other tables used in the
application will be introduced in later chapters when they are needed.
As previously mentioned, a new
logs
table is created every day. This means
that you can have a lot of tables accumulating in the logs datas, and often
the old ones aren't interesting any more. To avoid having too many tables
that you're unlikely to care about (and unlikely to want to pay for), we set
the table expiration time to 30 days in the future when the table is created.
This means that after a month, the tables will expire and you won't have to
worry about seeing them or paying for them any longer.
The sensor data table has all of the sensor and application information that
has been streamed in from mobile devices. It has a very wide schema. One
advantage of BigQuery's columnar storage system is that empty fields are
“free”—they are free to store, and they cost money only in queries when they
are explicitly used. So you can add lots of fields that may be used only in
limited circumstances. We won't go into detail about all of fields, but
Table
4.2
describes some of the more interesting data points collected from the
mobile devices.
